Color adaptation appears under two different circumstances: Temporal adaptation (also known as chromatic adaptation) results from the aftereffect of a colored stimulus on another stimulus presented to the same retinal area, for example when the illumination of a scene changes. Its effect can be observed by examining a white object like a piece of paper under different types of illumination, such as daylight, fluorescent or incandescent.
Spatial adaptation results from the influence of a relatively large colored stimulus (called context or surround) on a small stimulus (called infield) contained within it: By varying the surround, the (perceived) color of the infield changes.
The same physical stimulus, characterized by its spectrum, looks different within a changed context; on the other hand, different physical stimuli may look the same within differnt context. The second effect is the well known as color-constancy: Although the spectrum of the light reflected by well known objects differs depending on their illumination, their color ``looks the same''.
The experimental paradigm of cross-context-matching (see Krantz, 1968) can be utilized to investigate the effects of temporal as well as spatial adaptation. The subjects are presented an infield-surround-configuration. The stimuli within one context shall be changed until they look the same as a standard within another context: In such experiments the subjects are instructed to vary the color of a stimulus B surrounded by context T by adjusting some or all of its color-attributes until that stimulus appears to have the same color as a target stimulus A presented within another context S (see figure 1). The two infield-surround-configurations may be presented either simultaneously or successively to the subjects.
Figure 1:
Cross-context matching: The context S contains the target stimulus A. It is the subjects task to manipulate stimulus B presented within context T until it looks the same as stimulus A. The stimulus-context-configurations can be presented either simultaneously or successively.
If the colors are presented successively, the color of the first infield has to be remembered. Newhall, Burnham and Clark (1957) report higher variability of such memory matches and less time for their production. Furthermore, the stimuli reproduced from memory exhibit higher saturation and a slightly higher luminance than the standards originally presented. Hering (1920) coined the term memory color for these colors reproduced from memory. Katz (1930) suggests that these are related to the colors of well known objects. Memory colors seem to be distorted away from the original color to so called focal colors (cf. Heider, 1972), which are prototypical and especially easily remembered by members of different cultures speaking different languages. Those focal colors are highly saturated ones.
Colors can be characterized by a small number of attributes, such as hue, brightness, and saturation. Formally they also can be represented as elements of a convex cone embedded into the three-dimensional real vector space (Graßmann, 1853, elaborated by Krantz, 1975). Thus a single color can be denoted by three real numbers (whose value depends on the special coordinate system chosen - called color space - such as the CIE-chromaticity diagram).
Since changing the temporal or spatial context results in a change of the colors within that context, the effect of a change in context can also be modeled by a transformation of the coordinates representing the stimulus within the different contexts.
The goal of the study presented here is to investigate which kind of transformation describes the change of a color presented within two different contexts. Cross-context matches shall be predicted by a transformation that maps the color-space onto itself. the following three kinds of mappings are investigated (see also table 1):
Table 1:
Different kinds of transformation by adaptation
Figure 2:
Context-effect predicted by an affine transformation:
The effect of the change from a context C (standard illumnant C) to a context A (standard illuminant A) can be represented by an arrow starting at the color coordinates of the target stimulus within the original context and ending at the color coordinates of a stimulus having the same color within the changed context. The parameters of the transformation shown here are estimated by Burnham et al. (1957) from matches between illuminant A and C empirically established using not the stimuli shown here. It is obvious that the transformation setimated by Burnham et al. (1957) maps some stimuli to coordinates lying outside the convex cone representing all colors.
To differentiate between these three models a study is conducted using carefully selected stimuli: Assume a linear or affine transformation mapping the color presented within one context onto the same-looking colors produced by the subjects within a different context. For some highly saturated colors there should exist no matching color in the changed context under this condition, since the transformed color would come to lie outside the boundaries of the convex cone representing all colors (see figure 2). Only a projective transformation is capable of mapping highly saturated colors lying at the boundary of the convex cone onto that boundary again.